That bird link https://www.inaturalist.org/observations/identify?iconic_taxa=Aves has almot 55k pages (i.e. 1.6M obs). Adding lrank=class reduces them to 10%. I looked at a few of the finely ID’d ones, most seem to have been done directly by the observer (this might be a singular case because birders upload lots of obs - idk)…
There is a small residual risk when IDing something because it looks right (even perfectly right). The harder part is to know (or guess) which other things outside one’s knowledge also look almost the same but aren’t related. E.g. I mis-ID’d some beetle as Caraboidea (Ground beetles, Tiger beetles and allies) because it looked like one and was corrected that it was a Malachite beetle, a visually similar but unrelated group I have never heard about before. https://www.inaturalist.org/observations/163348255.
An observation identified as “Kingdom Plantae” or as “Kingdom Animalia” is a candidate to any of the 1,000 projects. The choice of the destination project depends only on the computer vision suggestions. For instance, an observation identified as “Kingdom Animalia” may be pushed to the project “Unknown / Cactaceae” (if and only if the CV suggestions indicate a cactus).
I provide a CV-based filter and iNat provides an ID-based filter.
They are 2 independent and complementary tools.
It is up to you to combine complementary tools, for instance for filtering out observations identified as animals when reviewing the project “Unknown / Cactaceae” (should there be much too many animals in the project for cacti).
… and AIUI, the CV suggestions should only include matches from the same iconic taxa group as the current ID.
So even if your project logic doesn’t care to filter those out as ID’s are improved, or to vacuum out stale suggestions from old CV models, anything already with a correct coarse ID will not be misidentified by the CV to another taxa above that rank.
Hence my assertion right from the very beginning that correct coarse IDs will help, and should never hinder, both people using your projects and people not using them.
Arguing about the proportions is irrelevant, that’s going to vary by day, and by species, and by observers - all that matters is that the arrow of change is always pointing in the positive direction when someone adds the best ID they are able to make correctly.
As @environ just stated, the CV would never suggest Cactaceae on an observation already identified as Kingdom Animalia. It would only suggest child taxa of Kingdom Animalia.
That is a very low bar. I also want to make good and effective use of my iNat time. I want to learn about biodiversity. Plant plant plant does not do it for me.
This suggests that I might not wait 30 days before an observation is a candidate to the phylogenetic projects: better collect the CV suggestions ASAP, before there is a correct or wrong ID, as my purpose is to provide a CV-based filter, not to mix it with the concept of ID-based filter. (Of course iNat has a different perspective, as it does not provide the CV suggestions for filtering). Or I don’t change anything, assuming coarse IDs (in fact, Community IDs) are most often correct.
I tried with a photo of an Opuntia, with location S.E. France, where we do have some cacti. Suggestions obtained before/after IDing the observation with a wrong ID, a species of green parrot that is exotic and invasive in Europe. Vision scores in brackets, range 0-100:
Sure, nobody is saying “Drop everything, because marking every unknown as Plant or Animal is the most valuable thing you can do”.
Just that if you see one, and you do that, at best you’re doing Some Good, and at worst you’re doing No Harm. And there’s not many things in life that are as cut and dried as that! :D
I’d generally think they’re more often at least in the right ballpark than the CV working completely blindly.
If your overall goal is to collate things in ways that are useful for helping specialists ID them finely, then it seems to me you are better off having things that are grasses appear among grasses because someone ID’d them as Plant, than having them lost among the Animals and only seen by people who don’t ever ID grasses.
I pointed out the problem from the perspective of false positives, but each false positive is equally a false negative by not appearing in the groups where people who might recognise it are looking.
That is completely fine. Plant plant plant might do it for somebody else though. And if they are happy to contribute in such a way, that’s great.
Of course, their coarse IDs will in theory always be of less value to the platform than those of an expert identifier who is the only one IDing a certain taxon they also happened to have written their phd thesis on at one point. But they are still helpful overall, so IMO we shouldn’t discourage it.
How you can spend your iNat time most effectively is up to you and no one should dictate that.
I usually take several pictures, trying to show all of the features like top/bottom of a leaf, flowers, stem. Animals are harder if they refuse to stay still. If the cv suggests the same thing for multiple pictures, I’ll use it’s suggestion for a higher level group. Don’t ask me if it’s family, genus, tribe, I have no idea. After that, I leave it up to the community to figure out. I won’t change it not because I’m stubborn, but because I am completely unqualified to agree or disagree.
I would never invest so much time to develop a solution providing a gain factor of only 2. (If this is what it is about). I don’t care whether the gain factor is 50, 100 or 200 (except if it’s very easy to improve it).
Conclusion based on a wrong hypothesis.
Clarification made meanwhile, so it’s OK now.
Thank you @arboretum_amy for the help.
These 2 propositions look fine, but their combination… not as expected:
I would like to learn more, so I identify “that tree on the left” as Dicot (just as a way to follow that observation) and wait for a clearer ID to come. Later an identifier (with expertise on “that tree on the left”) comes, but falls into this situation:
Then?
So, in the first place, the issue is all about using a placeholder (instead of a note or comment) but erasing placeholders (that ought not to be filled, to begin with) (except for using the placeholder as a clipboard, see below) is not an issue. I can’t agree, for a practical reason.
Extrapolation about someone else rationale or objective.
We absolutely do not intend to assign a role to the placeholder.
Our rationale is:
half of recent placeholders in S.A. contain names of species of genera,
valuable data shall not be lost,
consequently placeholders need be saved and kept visible.
You think some observers don’t do things in the right way. Yet, it is not an accident: iNat’s incentive produces this observable effect, continuously over time. In such a situation, it is not a good response (here I don’t mean your response, I mean iNat’s response, or lack of response) to consider that nothing is to be changed, just because there exists in principle another way to do it (by using a note or a comment). Yet, no statement (“better use a note or comment”) can solve a practical issue, only an action can: either removing the feature placeholder completely (so that all users will use a note or comment), or saving it and keeping it visible.
Inexperienced observers follow iNat’s incentive about the placeholder. But that way is wrong, and with experience the observer learns to use a note or comment instead. Satisfactory? Yes, because anyway the observer is still there (likely) and, on request, they will repeat the lost info? Provided they didn’t get it lost either, depending on what it was about. Satisfactory? I don’t think so.
The placeholder is just like a clipboard after the action “copy”. Don’t forget to “paste”, or the data will be lost! This “paste” should be automated: iNat should save the placeholder in a note or comment. That would be satisfactory. Saying that, I don’t force the placeholder to take a role. iNat does, and goes on, in an unsatisfactory way.
Since iNat first ‘tricks and bullies’ observers into using a placeholder, the first identifier is ‘actively complicit in destroying’ the observer’s data. That. Does. Harm. First.
The ID stands on a tripod. If iNat and the observer tangle in placeholder, then the identifier has to be the adult with whom the IDs stands. Or falls.
Today I saw an ID from a botanist, to family. The placeholder they missed was for the common name, tho slightly garbled - but an identifier could use the placeholder to go straight to sp. Could, and I did thanks to JP retaining the placeholder for me.
…and just when I thought we’d stopped talking past each other and had turned a corner toward a genuine two-state solution …
I’m only replying to this to leave you no doubt that any misrepresentation was not part of some Machiavellian master plan to malign or ‘bully’ you.
If I’d intended it to be a direct quote I’d have used the forum quoting mechanism to link to where you’d said it. I thought it should have been obvious that “I was paraphrasing” a concept I thought we both understood. There evidently seems to be a language barrier, and the only way I know to cross those is to find pieces of common ground to anchor ideas to.
You actually said:
Which is not quite how I’d remembered the takeaway of it from when I’d first read it - but I clearly wasn’t trying to lure you into some logical trap that hinged on exactly the differences between both these sentiments either.
I apologise for any offense in this. None whatsoever was intended.
That doesn’t change how much I still stand by that rationale even if you refuse being associated with it. Practical optimisation is almost always the sum of many small gains made in many places. Actions that can only have a positive effect on what is being optimised for, which are cheap or opportunistic, are a valuable part of that.
The time invested by one person correctly clicking Plant in just one observation can have an enormous multiplier gain if you consider how many other identifiers it may save from having to consider, then scroll past, something they know they can’t be any significant help with.
You may say “but they wasted seconds to save milliseconds!!” the proportion is tiny. But milliseconds multiplied by millions of repeated actions quickly add up to a lot of time wasted and a lot less work done.
Naive faux-pas are not “active complicity”. And if I was one of the people who had accidentally done this, or the inat developer who created a feature that you continue to vehemently insist on using in ways it apparently wasn’t intended - I’d probably be feeling pretty bullied by this continual, somewhat hyperbolic, onslaught.
As I noted previously, any problems with placeholders are completely independent from the issue of whether or not adding broad IDs is beneficial. There are an uncountable number of ways that an inexperienced user can make mistakes that someone more experienced will need to correct. If we don’t trust them they won’t make mistakes, if they don’t make mistakes they won’t learn.
If the only place you kept an important record of something that there would be terrible consequences to losing was in the cloud on inat, then that’s your mistake to learn from. There are an uncountable number of ways that any data here could, without warning, vanish without a trace or any certain hope of recovery.
I’m clearly missing something relevant in what you’re trying to say there. If a botanist could only comfortably ID to family, why should we trust any more specific ID that couldn’t be made (and wasn’t in a later edit made by the original observer), without the aid of the placeholder?
But I didn’t come here to argue, or to try and convince other people to do what I do the way I do it without thinking for themselves. If I want a conversation where I have to find a way to share some inconvenient but important insight with someone who loudly and with absolute unmovable certainty insists their way is the only acceptable way, I … have a cat. I really don’t need to come here to play that game …
I’m here to share what I’ve seen and learn from what others have noticed before me, not to pick a side in someone else’s holy war.
If most people tell you “that’s a hammer, it’s not really the best tool for putting in screws” - but you keep hammering them in and don’t tell others making the same mistake about the existence of the screwdriver set, then I’m not going to mistake myself for the person who can Finally Convince You.
It doesn’t mean I don’t appreciate the other Genuinely Good Work you do, but I can’t ease your pain on this one and won’t pile on to bullying the developers to make changes they’ve considered and, for whatever reason, rejected. Reality is what doesn’t go away when you ignore it - and placeholders not being what you would really like them to be appears to be reality. Teach people to use notes or comments for things they want to be persistent and not vulnerable to edits from the broader community is the only advice I have on that.
So, it is actually not about whether I matter “whether it’s 2 or 100”.
I would reformulate as “whether it’s 50 or 100”.
How could we guess “2 or 100” means “50 or 100”?
Please be precise from the start. Thanks for the clarification.
Yes, there is an impact. In short: a coarse ID hidding/erasing the placeholder makes the community lose the finer ID that was in the placeholder (50% of all placeholders are IDs at rank species or genus). Only an ID as precise as the placeholder would be neutral, with regard to the placeholder, not a coarse ID.
Could we agree at least about the facts?
(Only then come the opinions, about whether or not a change or action is unnecessary, desirable or needed).
Last attempt. I had an earlier post on the forum - no one wants to use a disappearing placeholder. I don’t know why iNat offers it.
observer has an ID to species - problems, tiny typo, missing species, poor internet.
identifier adds a broad ID.
If the obs were straight forward, the observer’s sp would be Needs ID, and the identifier’s ID would be ‘support’. Working both as intended and as expected by all parties.
As iNat is working as intended, we have the broad ID at Needs ID, and iNat dismisses the ID from the observer. End of.
What purpose does the placeholder serve? Why is it there? At all ? It should not be possible for a good intentioned ID to destroy useful information. Like a bulldozer crushing counterfeits
Yes. We (non-experts) are constantly being told in other topics that there is such a dearth of identifiers that we should all be identifying. We should identify twice as much as we observe. At any level we are comfortable with. Because even coarse IDs can help, and it will help us to learn.
